Re: [PATCH v4 5/8] transport: add client side capability to request object-info

[Jonathan Tan <jonathantanmy@google.com>]

Calvin Wan <calvinwan@google.com> writes:
> Sometimes it is useful to get information about an object without having
> to download it completely. The server logic has already been implemented
> as “a2ba162cda (object-info: support for retrieving object info,
> 2021-04-20)”. This patch implements the client option for it. Currently,
> only 'size' is supported and the server must be v2, however future
> patches can implement additional options.
> 
> The question of version mismatch often comes up with client/server
> relationships. There are two cases to consider here (assuming either
> client or server functionality for object-info changes between the
> different versions):
> 
>  1. client version > server version
>  	- client can request additional attributes from server
>  2. server version > client version
>  	- server can return additional info to the client
> 
> The second case is a non-issue since the client would never be able to
> request additional info from the server.  In order to solve the first
> case, recall an earlier patch where the server sends back the attributes
> even if no object ids are sent by the client. This allows the client to
> first check whether the server can accept the requested attributes
> before sending the entire request.

From this description, it seems like the intention is to send an
object-info request, and then if the server responds in a certain way
(here, sending back the attributes - presumably without sending any
actual information), then we know that the server doesn't support our
request and we need to fall back. As Junio says [1], this requires one
RTT more than necessary. We could just determine if the server supports
the attributes we want by using capabilities (without needing to send
the request to check).

[1] https://lore.kernel.org/git/xmqqilqnsaep.fsf@gitster.g/